Nano porous silicon microcavity sensor for determination organic solvents and pesticide in water

The basic characteristics of the porous silicon microcavity (PSM) and the resonant wavelength shift (Δλ) caused by the ambient refractive index (n) were determined experimentally by using a series of liquids with known refractive indices. The effective refractive index of the nano-porous silicon layer immersed into organic solvent would be increased due to the substitution of air with liquid in the pores and consequently the optical thickness of the layer is increased. As a result, the resonant wavelength shift would be dependent upon the refractive index value of the organic solvent. Table 1 presents a series of organic solvents such as methanol 99.5%, ethanol 99.7%, isopropanol 99.7% and methylene chloride 99.5% (product of NHTC-China) with their refractive index and resonant wavelength of the sensors, when they are dipped into corresponding organic solvent for some minutes.

Sensitivity (Δλ/Δn) is one of the most important parameters for evaluating the performance of the sensors. Using the experimental data in table 1, we calculate the sensor sensitivity of about 200 nm RIU⁻¹. The Spectrophotometer Varian Cary 5000 is able to detect a wavelength shift of 0.1 nm, and then the minimum detectable refractive index change in the porous silicon layer is less than 10⁻³. Experiment shows that after complete evaporation of organic solvent, the reflectance spectra of the sensors return to their original waveform position (as in the air). In our case the evaporation of organic solvents in open air at room temperature was completed for 40–50 min, but this process can occur for 20 s when the samples were in the vacuum chamber with 10⁻¹ torr. That means, the change of sensor reflectance spectra are temporary and it is useful for reversible optical sensing.

An important parameter of the microcavity sensor is the change of the refractive index of the porous layer. It depends on the refractive index of the liquid as well as the porosity of the porous layer. In Bruggeman effective medium approximation, the relation between effective refractive index of the pore layers (${{n}_{PSi}}$), silicon refractive index (n_si = 3.5), void refractive index (n_void = 1) and porosity (P) is presented by following equation

$$\begin{eqnarray}&&\left( 1-P \right)\frac{{{n}^{2}}_{Si}-{{n}^{2}}_{PSi}}{{{n}^{2}}_{Si}+2{{n}^{2}}_{PSi}}+P\frac{{{n}^{2}}_{void}-{{n}^{2}}_{PSi}}{{{n}^{2}}_{void}+2{{n}^{2}}_{PSi}}=0.\end{eqnarray} \tag{ 1 }$$

Based on this relation we determined the porosity (P) of the pore layer and the refractive index of the pore layer (${{n}_{PSi}}$) dipped in liquid with known refractive index. The simulation calculations for basic characteristics of sensor consist of the following steps:

• (i)  
  Determination of refractive index of the defect layer based on the experimental resonant wavelength and the pore layer thickness from scanning electron microscopy (SEM) image.
• (ii)  
  Determination of layer porosity in the air using the relation (1). When the refractive indices of the pore layers in air changed from 2.739 to 1.323, the porosity of the layer changed from 30 to 80 by calculation.
• (iii)  
  Determination of effective refractive index of layer when the voids were filled by liquid with known refractive index using relation (1).
• (iv)  
  Determination of resonant wavelength of sensor dipped into liquid using the simulation reflectance spectra.

Simulation shows that the contrast of the porosities (i.e. refractive indices) of the layers strongly influences the wavelength shift (i.e. the sensitivity) of the microcavity. The contrast of the porosities would be high when the change of current density was large in the electrochemical etching process. However, experiment shows that the imperfection of the interfaces of layers increased with the large change of current densities. In our work, when the porosity contrast of layers is more than 40, the reflective spectra of the device are deformed in the reflection intensity and the line-width of transmittance zone.

The curves from C1 to C4 in figure 4 present the fitting process of sensor basic characteristics by simulations with that by experiment (curve E). The fitting showed that the porosity contrast between two layers affects the sensor sensitivity (Δλ/Δn). Consequently, the matching process found suitable porosity of 34% and 72% of low and high porosity layers of the prepared sensor, respectively.

Zoom In Zoom Out Reset image size

The microcavity-based sensors have been applied to determination of different solutions of ethanol and methanol in the commercial gasoline A92.

Figure 5 shows the measured results of the resonant wavelength shift of the microcavity sensor immersed into gasoline A92 with different concentrations of ethanol and methanol. In the case of a mixture of ethanol/A92, a resonant wavelength shift is of 3.6 nm, when ethanol concentration changed in the range from 5% to 15% in the gasoline. With the sensitivity of the sensor as described above, the minimum determination of ethanol concentration change in the gasoline is of about 0.4%. In the case of methanol/A92, wavelength shifts are of 7.2 nm between the 5% and 15% methanol mixtures, respectively. From these experimental data, we suppose that the elaborated sensor can distinguish change of about 0.2% in concentration of methanol in the gasoline.

Zoom In Zoom Out Reset image size

It is known that the response of the sensor depends on solvent vapor pressure in the sensor chamber [17]. This vapor pressure is related to the vapor pressure of the solvent in the solution chamber through a gas stream flowing through the solution. Assuming that the vapor pressure in the solution chamber obeys the rules of vapor pressure in a closed system, the relation between the wavelength shift (Δλ), the vapor pressure in the solution chamber (P_solution) and the velocity of airflow (V) is crudely presented as $\Delta \lambda \tilde{\ }{{P}_{solution}}\;\theta (V)$, where $\theta (V)$is an empirical function of V, which shows dependence of concentration of solution on the velocity of airflow. The P_solution can be calculated by the following formulae [18]

$$\begin{eqnarray}&&{{P}_{solution}}=\mathop \sum \limits_{i} \left( {{P}_{i}}{{X}_{i}} \right)\;\;\left( {\rm Raoul's}\;{\rm law} \right),\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{{{\rm log} }_{10}}{{P}_{i}}=A-\frac{B}{C+T}\;\;\left( {\rm Antoine}\;{\rm equation} \right),\end{eqnarray} \tag{ 3 }$$

where P_i is the vapor pressure of a particular substance, X_i is the corresponding mole fraction of that substance, A, B and C are component-specific constants (the coefficients of Antoine's equation), 'i' is an indexing component that keeps track of each substance in the solution.

Equations (2) and (3) show that Δλ is a function of V, X_i and P_i(T). Below we consider those relations in the experiment. We carried out experiments on ethanol and acetone solution. These are very common organic solvents and some of their physical properties such as boiling point, refractive indices and Antoine's coefficients from [18] are shown in table 2.

Figure 6 shows the dependence of Δλ on T, Δλ(T), for acetone and ethanol solutions with various concentrations at the airflow velocity of 0.84 ml s⁻¹. Equations (2) and (3) show that we can consider the temperature dependence of Δλ(T) as the temperature dependence of P_i(T) modified by multipliers X_i if it was assumed that the contribution of water to solution pressure is small. Experimental data from curve 1, which describes Δλ(T) of the water, shows the validity of this assumption in our measurement. P_i(T) steadily increases as temperature increases (i.e. a monotonically increasing function), so the curves of solvent solutions with various concentrations are separate; for example, the curves 2–4 of ethanol solutions or the curves 5–7 of acetone solutions. Using equation (3) we calculated the rate of change of P_i(T) for acetone and ethanol in the range of temperature from 30 °C to 50 °C and its values are presented in table 2.

Zoom In Zoom Out Reset image size

The slope of P_acetone (T) is greater than that of P_ethanol (T) in the studied range of temperature (see table 1), so the curves describing Δλ(T) of acetone and ethanol solutions are intersecting with each other not more than one time (for example: curves 3 and 5), or not intersecting (for example: curves 3 and 4). Consequently, a curve describing Δλ(T) characterizes the solution of acetone (or ethanol) at a given concentration. In other words, the dependence of the wavelength shift on the solution temperature discriminates between solutions of ethanol and acetone with various concentrations [19].

Figure 7 shows the dependence of the resonant wavelength shift Δλ(C) on ethanol concentration, when velocity of the airflow (V) and temperature of the solution (T) work as parameters in the measurements. It can be seen in figure 7 that the curve described by Δλ(C) is linear and its slope, i.e. sensitivity of the measurement, increases as V and T increase. These remarks are also deduced from equations (2) and (3) when X_i is a variable, and T and V are parameters. Linearity of this dependence is a favorable condition for determination of solvent concentration. The increase of the slope creates an increase in sensitivity received from the measurement. From data in curves 2 and 3, which were received from the measurement with parameters T and V at 45 °C and 0.84 ml s⁻¹, and at 30 °C and 1.68 ml s⁻¹, we obtain the difference of Δλ of about 18.5 nm and 10.0 nm, respectively, between 0% and 100% ethanol. However, having measured with this sensor in the liquid phase in this concentration range, we obtained the Δλ difference of about 5 nm only [20]. Therefore, the sensitivity received from the measurement in the vapor phase with the value of T and V at 45 °C and 0.84 ml s⁻¹, and at 30 °C and 1.68 ml s⁻¹ increases 3.7 and 2.0 times, respectively, as compared with that in the liquid phase. We expect that the sensitivity of the measurement can be strongly improved with a reasonable combination of both parameters T and V.

Zoom In Zoom Out Reset image size

Figure 8 shows the dependence of Δλ on V, Δλ(V), at a temperature of 30 °C when concentration of ethanol and acetone work as the parameters. It can be seen in figure 6 that curves describing Δλ(V) are separate straight lines with different concentrations of acetone and ethanol. This shows that empirical function θ(V) is a linear function of V. Now, we consider properties of the slopes of curves in figure 8. According to equation (2), the slope of the curve describing Δλ(V) increases as P_i and X_i increase. We apply the obtained results for curves 2 and 3 received from measurements with ethanol and acetone solutions at the same concentration (20%). It can be seen that the vapor pressure of acetone is larger than that of ethanol (see table 2), so the slope of curve 3 is larger than that of curve 2. We also apply the results for curve 2 and 4 received from measurements with ethanol concentrations at 40% and 20%. The slope of curve 4 is larger than that of curve 2 due to the greater value in the concentration.

Zoom In Zoom Out Reset image size

It is deduced from figure 8 that dependence of the wavelength shift on velocity of the airflow is linear, and the slopes Δλ/ΔV are 2.4 nm ml⁻¹ s⁻¹ and 3.7 nm ml⁻¹ s⁻¹ for the same concentration of 20% ethanol and acetone solutions, respectively. In addition, when the concentration of organic solvent increases, the slopes Δλ/ΔV would be enhanced (for example, the value of Δλ/ΔV enhanced from 2.4 nm ml^−1s to 3.4 and 5.1 nm ml⁻¹ s⁻¹ when the concentration of ethanol increased from 20% to 30% and 40%, respectively). Based on this phenomenon, we can simultaneously determine the kind and concentration of organic content in the solutions. For example, 40% ethanol and 20% acetone have similar temperature dependence (see figure 6) but can be discriminated by their air flow velocity dependence, and while 30% ethanol and 20% acetone have similar air flow velocity dependence (as can be seen from figure 8), they can be discriminated by their temperature dependence.

The very low concentration of atrazine solutions was obtained by dilution from mixture, obtained by stirring 21.5 mg of atrazine in 1000 ml of ultrapure water and a minimum amount of ethanol to ensure solubility (concentration of atrazine was of about 21.5 ppm or 10⁻⁴ M). To determine very low concentrations of atrazine solutions in the range between 2.15 and 2.15 × 10⁶ pg ml⁻¹ (from 10⁻¹¹ to 10⁻⁴ M), we have measured the cavity-resonant wavelength shift of the nano porous silicon microcavity sensors with various conditions: atrazine in pure water and in an aqueous solution of a humic acid (HA, 0.2 mg ml⁻¹) extracted and purified from soil. Humic acid solutions were chosen to represent systems similar to natural conditions where water-containing pesticides also dissolve organic matter as component [21]. When an atrazine solution dropped on to sensor surface, the solution would partially substitute the air in the pores of each layer of sensor device caused a change of its refractive index. We observed a repeatable completely reversible change in the cavity reflectivity spectrum.

To test the performance of the optical sensor for the determination of atrazine pesticide, we studied the wavelength shift in the reflectance spectra with various conditions: in air, in pure water and in humic acid (HA). The effective refractive index of the nano-PSMC layer immersed into solutions would be increased due to the substitution of air with liquid in the pores, and consequently the optical thickness of the layer is increased. When the microcavity sensor was exposed to water (with refractive index of 1.3326) and to humic acid (with refractive index of 1.3541), the reflectance spectra promptly shifted towards longer wavelengths by about 39.2 nm and 46.5 nm, respectively.

After analyzing the resonant wavelength shift in the reflectance spectra of microcavity sensor in various conditions, we performed the wavelength shift measurements for the determination of atrazine pesticide in water and HA during their exposure to different concentrations (2.15–2.15 × 10⁶ pg ml⁻¹). It is remarkable that the sensor response depends mainly on two physical factors: the refractive index of the atrazine solution (concentration of atrzine) and its capability of filling the PS pores. The concentration of atrazine in the solution is determined by the wavelength peak shift of the sensor, and the capability of filling the pores is tested by the repetition of measurement values. As shown by experimental results, the resonant peak shift in the reflectance spectra of 1D-PSMC structures for different atrazine concentrations in water from 2.15 to 2.15 × 10⁶ pg ml⁻¹ is of 21.1 nm, but the sensor responses are non-linear in a large range of atrazine concentration. Figure 9 presents the response curve of the sensor to atrazine in water with concentration from 2.15 to 2.15 × 10⁶ pg ml⁻¹. The wavelength shift is only linearly increased in the very low concentration of atrazine (from 2.15 to 21.5 pg ml⁻¹) in our measurement.

Zoom In Zoom Out Reset image size

Table 3 presents the measurement results of resonant wavelength shift of sensor wetted by atrazine solutions with low pesticide concentrations. The resonant wavelength of sensor shifted on 6.7 nm and 12.3 nm when the concentration of atrazine changed from 2.15 to 21.5 pg ml⁻¹ in water and in humic acid, respectively. It is an important factor for sensor applications that wavelength shift versus atrazine concentration in very low range is linear. Figure 10 shows a linear relation between the different concentrations of atrazine in the very low concentration range from 2.15 to 21.5 pg ml⁻¹ and the resonant peak wavelength shift. In figure 10, each experimental point was the average value of five independent measurements, with the accuracy representing the standard deviation. We could calculate the sensitivity of the sensor as the slope of the linear curve interpolating the experimental points.

Zoom In Zoom Out Reset image size

Thus, we obtained the sensor sensitivity value of 0.3 and 0.6 nm pg⁻¹ ml⁻¹ for atrazine in water and in humic acid solution, respectively. From these measurement results, we also estimated the limit of detection (LOD), as the ratio between the instrument resolution and sensitivity. LOD numerical value is 1.4 and 0.8 pg ml⁻¹ for atrazine in water and in humic acid solution, respectively. In addition, it was observed that the higher wavelength shift was observed in the case of atrazine in HA, because atrazine with HA contains dissolved organic matter as the component having higher refractive index in comparison with water.

It is remarkable that the sensor fabricated by our method has significant improvement for the determination of pesticide present in water in comparison with previous works (for example with [12] and [21]). It may be caused by different current densities and etching times for preparation of microcavity samples (i.e. difference in the porosity ratio of low- and high-refractive index layers and layer thickness) and by difference of cavity resonant wavelengths (visible versus infrared). In our case, the experiment had been done for several measurements and the results have good repetition. On the other hand, the obtained results were checked by comparison with electrochemical immunoassays [22] with the same method for preparation of low concentration atrazine sample. In addition, it was observed that, after moving the atrazine solution on the sensor surface and washing it with distilled water, the cavity-resonant wavelength in the reflectance spectra promptly returns to its original position. This is a very good quality of these structures, as it is helpful in the development of reversible sensing devices.